Adam Rice

I like the `?` and `!` but there's no doubt they're super confusing to people coming to the spec for the first time. Would we have to explicitly write "Assert:...

For Blink's implementation, knowing whether something can throw is useful, because it changes the calling convention. However, in practice there's stuff that can throw in Blink that never throws in...

What happens when we call an operation that could throw, but which is annotated with a "!" to indicate that it doesn't throw in this case, is that we turn...

Chrome is interested.

I realised that for synchronous iterators we could get optimum performance if we just enqueued everything in startAlgorithm. It's probably not worth the extra complexity, and I guess an iterator...

Thanks for the feedback. I have a suspicion that it's detectable, but I don't know how. I'm ready to land this if you are.

Ah, wait, we need interest from another implementer. @youennf, @evilpie, any thoughts?

That's really interesting. I don't like either option, but I'm leaning towards special-casing arrays. My reasoning is that I want people to use the `ReadableStream.from([chunk]).pipeTo(writable)` pattern, but if it performs...

It depends on your definition of "serious performance benefits". I went away and implemented it, and in a microbenchmark ``` ReadableStream.from([chunk]).pipeTo(writable); ``` takes 83μs with the array optimisation and 139μs...

To be clear, I was only trying to evaluate the performance of treating Array as a special case, I was not trying to make the behaviour match the async iterator...